본문 바로가기
  • Home

A Practical Approach to Fall Detection: Lightweight AI for Low-Cost Implementation

  • Journal of The Korea Society of Computer and Information
  • Abbr : JKSCI
  • 2025, 30(1), pp.53-63
  • Publisher : The Korean Society Of Computer And Information
  • Research Area : Engineering > Computer Science
  • Received : December 6, 2024
  • Accepted : December 27, 2024
  • Published : January 31, 2025

Minhyung Ryu 1 Jaegoo Shim 2

1니어네트웍스
2대구보건대학교

Accredited

ABSTRACT

This paper proposes the development of a lightweight AI-based fall detection system that can operate efficiently in low-spec hardware environments. Existing fall detection research has often relied on high-performance GPU environments or focused solely on detection accuracy, without adequately addressing processing speed and hardware limitations. To overcome these challenges, this study utilizes MediaPipe, a GPU-independent framework, combined with data analysis-based machine learning models, to design a system that requires minimal computational resources while ensuring fast and accurate fall detection. Experimental results demonstrate that the proposed model outperforms GPU-based models in terms of training and inference speed, while also delivering competitive detection performance. This research presents a promising approach to reducing dependency on high-cost hardware, thereby maximizing the practicality of fall detection systems in real-world scenarios.

Citation status

* References for papers published after 2023 are currently being built.